These cheaper front end servers handle many of the requests, leaving the expensive (in terms of compute) jobs to the back-end application servers. Delegating the easy content to the edge servers, mean you require less expensive application servers, thus, you save money and increase response times to your users. Server instance sizing
You don’t have to have the same size server (in terms of compute and memory) which enables appropriate sizing of the server to be deployed, depending on the workload. As mentioned in the caching section, workloads can be different between the frontend and back end servers, and based on scaling and workload, can enable optimisation of the application performance as well as costs.
Often a greater number of smaller servers, gives you flexibility to follow the workload closely rather than having few larger servers. Of course, if the demand is relatively static that may suit your application better.
This is where monitoring can show dividends regarding the utilisation of resources such as server CPU and memory, allowing intelligent decisions to be made.
Reserved resources
AWS provide further billing options to reduce your costs. If you know roughly what your demand for resources is going to be, you can opt to pre-order vouchers to be used against charges for various AWS resources that will discount the monthly bill in exchange for an upfront commitment for 1 to 3 years.